Weaviate Tutorial

Combination of vector and keyword search for best results.

Installation¶

docker run -d -p 8080:8080 semitechnologies/weaviate:latest

Schema + Query¶

import weaviate
client = weaviate.Client('http://localhost:8080')

result = client.query.get('Article',['title','content'])\
    .with_hybrid(query='PostgreSQL indexes', alpha=0.5)\
    .with_limit(5).do()

Features: - Hybrid (vector+BM25) - Auto-vectorization - GraphQL API - Multi-tenancy

Architecture and Features¶

Weaviate uses its own vector index (HNSW) and combines it with an inverted index for keyword search. The alpha parameter in a hybrid query determines the ratio between vector similarity (alpha=1) and BM25 keyword relevance (alpha=0). A value of 0.5 is a good starting point.

Auto-vectorization means Weaviate can automatically generate embeddings when inserting data using built-in modules (text2vec-openai, text2vec-transformers). You do not need to manage an embedding pipeline externally. The GraphQL API enables flexible querying with filtering, aggregations, and cross-reference queries. For multi-tenant applications, Weaviate offers native tenant isolation with efficient resource utilization.

Weaviate for Hybrid¶

Best results thanks to vector + keyword combination.